An arXiv preprint presents a mathematical construction in which a smooth random flow on the three-torus is associated with exponential growth of a magnetic field. For each fixed sufficiently small resistivity, the theorem gives an almost-sure, all-time lower bound on the field’s L2 norm—the quantity used in the model to track its size—with a growth rate of at least one-half and a finite random prefactor. “Almost sure” means the statement holds outside an exceptional null set of random cases. The result concerns one specified solution of a linear resistive induction equation, not measurements from a physical experiment.
A result built around one flow
The paper asks whether a smooth, divergence-free random velocity field can display fast-dynamo behavior in that equation. Its model uses independent, uniformly distributed phase variables and a smooth velocity formula with cyclic indexing, evaluated at translated coordinates. The velocity is divergence-free and obeys deterministic pathwise bounds for all the time and spatial derivatives specified in the paper.
The field being followed is the random unique solution of the linear resistive induction equation on the three-torus, initialized with the paper’s specified cosine magnetic field. This makes the claim specific: the analysis concerns one mathematically defined construction and its associated solution, rather than a broad collection of random flows.
Turning randomness into a proof
The proof works in Fourier space, where the field is represented through modes. It follows three specially chosen Fourier modes and uses the random translations to produce a tridiagonal solution operator on selected rows. The argument then tracks how the resulting mode relations carry the lower bound through the construction.
At the key probabilistic step, Jensen’s formula is used to lower-bound the expected logarithm of a coefficient’s magnitude from the two extremal coefficients of a complex polynomial. Conditional expectations, exponential-moment estimates and a martingale bound control random fluctuations in the repeated lower-bound argument. The martingale lemma supplies a finite random variable that keeps that iteration under control.
The argument first establishes growth at discrete time steps. A Grönwall estimate—an inequality used to control how a quantity changes continuously—then transfers the mode bound to all real times. That step supports the theorem’s all-time conclusion rather than a claim only at selected discrete times.
The important qualification behind “almost sure”
The probability qualification is central. The exceptional null set may depend on resistivity, so the result handles each fixed sufficiently small value separately. It does not directly provide one deterministic realization that works almost surely for all small resistivities at once.
The random lower-bound constant is controlled in a different way. For some positive moment order, the corresponding moment of that constant is bounded uniformly over the stated small-resistivity interval. The theorem asserts that such an order and constant exist, but does not give their numerical values.
A narrow result, not a general recipe
The authors describe the construction as the first smooth random fast dynamo, while explicitly saying it is not a general theory. The result therefore remains within this specially constructed setting. Its tridiagonal Fourier structure does not establish the same behavior for generic smooth random velocity fields.
The initial condition narrows the conclusion further. The displayed evolution uses a specified cosine initial magnetic field, so the theorem does not establish exponential growth for every possible initial magnetic field.
That scope matters for how the work should be read. It is an analytic result about one random velocity-field construction and its resistive induction equation, relevant to mathematical dynamo theory and Fourier-based analysis rather than direct empirical evidence about physical dynamos.
An AI-generated idea, human-checked manuscript
The paper discloses that its central proof idea was generated autonomously by ChatGPT 5.6 Sol Ultra, while the author wrote and verified the manuscript. The disclosure identifies how the idea arose; it does not widen what the theorem establishes, which remains tied to the constructed flow and specified magnetic field.
The document is an arXiv v1 preprint dated 20 Aug 2026. The supplied text says the original and final generated TeX manuscripts are available in the arXiv source files. A broader theory covering classes of smooth random flows remains outside the result’s stated scope.
Paper data and sources
Original title: An AI-discovered smooth random fast dynamo on $\mathbb{T}^3$
Authors: Keefer Rowan
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text